]>
Commit | Line | Data |
---|---|---|
46abada0 BK |
1 | <?xml version="1.0" encoding="UTF-8" standalone="no"?> |
2 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> | |
4394b61e | 3 | <html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>Implementation</title><meta name="generator" content="DocBook XSL Stylesheets V1.73.2" /><meta name="keywords" content=" ISO C++ , library " /><meta name="keywords" content=" ISO C++ , library " /><link rel="start" href="../spine.html" title="The GNU C++ Library Documentation" /><link rel="up" href="concurrency.html" title="Chapter 40. Concurrency" /><link rel="prev" href="concurrency.html" title="Chapter 40. Concurrency" /><link rel="next" href="bk01pt12ch40s03.html" title="Use" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Implementation</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="concurrency.html">Prev</a> </td><th width="60%" align="center">Chapter 40. Concurrency</th><td width="20%" align="right"> <a accesskey="n" href="bk01pt12ch40s03.html">Next</a></td></tr></table><hr /></div><div class="sect1" lang="en" xml:lang="en"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="manual.ext.concurrency.impl"></a>Implementation</h2></div></div></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.concurrency.impl.atomic_fallbacks"></a>Using Builtin Atomic Functions</h3></div></div></div><p>The functions for atomic operations described above are either |
46abada0 BK |
4 | implemented via compiler intrinsics (if the underlying host is |
5 | capable) or by library fallbacks.</p><p>Compiler intrinsics (builtins) are always preferred. However, as | |
6 | the compiler builtins for atomics are not universally implemented, | |
7 | using them directly is problematic, and can result in undefined | |
8 | function calls. (An example of an undefined symbol from the use | |
9 | of <code class="code">__sync_fetch_and_add</code> on an unsupported host is a | |
10 | missing reference to <code class="code">__sync_fetch_and_add_4</code>.) | |
11 | </p><p>In addition, on some hosts the compiler intrinsics are enabled | |
12 | conditionally, via the <code class="code">-march</code> command line flag. This makes | |
13 | usage vary depending on the target hardware and the flags used during | |
14 | compile. | |
15 | </p><p> If builtins are possible, <code class="code">_GLIBCXX_ATOMIC_BUILTINS</code> | |
16 | will be defined. | |
17 | </p><p>For the following hosts, intrinsics are enabled by default. | |
18 | </p><div class="itemizedlist"><ul type="disc"><li><p>alpha</p></li><li><p>ia64</p></li><li><p>powerpc</p></li><li><p>s390</p></li></ul></div><p>For others, some form of <code class="code">-march</code> may work. On | |
19 | non-ancient x86 hardware, <code class="code">-march=native</code> usually does the | |
20 | trick.</p><p> For hosts without compiler intrinsics, but with capable | |
21 | hardware, hand-crafted assembly is selected. This is the case for the following hosts: | |
22 | </p><div class="itemizedlist"><ul type="disc"><li><p>cris</p></li><li><p>hppa</p></li><li><p>i386</p></li><li><p>i486</p></li><li><p>m48k</p></li><li><p>mips</p></li><li><p>sparc</p></li></ul></div><p>And for the rest, a simulated atomic lock via pthreads. | |
23 | </p><p> Detailed information about compiler intrinsics for atomic operations can be found in the GCC <a class="ulink" href="http://gcc.gnu.org/onlinedocs/gcc/Atomic-Builtins.html" target="_top"> documentation</a>. | |
24 | </p><p> More details on the library fallbacks from the porting <a class="ulink" href="http://gcc.gnu.org/onlinedocs/libstdc++/17_intro/porting.html#Thread%20safety" target="_top">section</a>. | |
4394b61e | 25 | </p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.concurrency.impl.thread"></a>Thread Abstraction</h3></div></div></div><p>A thin layer above IEEE 1003.1 (i.e. pthreads) is used to abstract |
46abada0 BK |
26 | the thread interface for GCC. This layer is called "gthread," and is |
27 | comprised of one header file that wraps the host's default thread layer with | |
28 | a POSIX-like interface. | |
29 | </p><p> The file <gthr-default.h> points to the deduced wrapper for | |
30 | the current host. In libstdc++ implementation files, | |
31 | <bits/gthr.h> is used to select the proper gthreads file. | |
32 | </p><p>Within libstdc++ sources, all calls to underlying thread functionality | |
33 | use this layer. More detail as to the specific interface can be found in the source <a class="ulink" href="http://gcc.gnu.org/onlinedocs/libstdc++/latest-doxygen/gthr_8h-source.html" target="_top">documentation</a>. | |
34 | </p><p>By design, the gthread layer is interoperable with the types, | |
35 | functions, and usage found in the usual <pthread.h> file, | |
36 | including <code class="code">pthread_t</code>, <code class="code">pthread_once_t</code>, <code class="code">pthread_create</code>, | |
37 | etc. | |
38 | </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="concurrency.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="concurrency.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="bk01pt12ch40s03.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Chapter 40. Concurrency </td><td width="20%" align="center"><a accesskey="h" href="../spine.html">Home</a></td><td width="40%" align="right" valign="top"> Use</td></tr></table></div></body></html> |